Aplicación móvil para clasificación automática de malas hierbas en imágenes usando aprendizaje profundo.
Weeds in the field, are considered a major problem for crops, cause negative effects that rival the corn plants in their development by removing all the nutrient elements, where the farmer uses any herbicide to eliminate these weeds by applying in all The crop For this reason, it is considered impor...
Zapisane w:
| 1. autor: | |
|---|---|
| Kolejni autorzy: | |
| Format: | bachelorThesis |
| Język: | spa |
| Wydane: |
2020
|
| Hasła przedmiotowe: | |
| Dostęp online: | http://repositorio.utc.edu.ec/handle/27000/6698 |
| Etykiety: |
Dodaj etykietę
Nie ma etykietki, Dołącz pierwszą etykiete!
|
Dodaj etykietę | Streszczenie: | Weeds in the field, are considered a major problem for crops, cause negative effects that rival the corn plants in their development by removing all the nutrient elements, where the farmer uses any herbicide to eliminate these weeds by applying in all The crop For this reason, it is considered important that this problem be treated. One of the solutions to mitigate this problem could be through the use of deep learning techniques that allow obtaining a mobile application with information to minimize or eliminate weeds in the corn crop based on the correct herbicide. The literature review allowed to determine similar studies for the classification of weeds in the crop through techniques used to classify automatically. However, the techniques proposed by the researchers have limited use of machine learning techniques. Therefore, a classification model based on deep learning techniques for the classification of weeds in images is proposed. As a result, a mobile application developed under the latest technology developed for the benefit of farmers, institutional herbarium, students and professors of related areas at the Technical University of Cotopaxi is obtained. The methodology used for the development of the research project was the Convolutionary Neural Networks, using the CRISP-DM methodology that corresponds to the description of the phases. In addition, the Android Studio programming language and the Custum Vision platform will be used allowing learning algorithms to be used, with the Tensorflow open source library capable of building neural networks, used for the detection of people, animals, plants and places. The generated model was evaluated achieving an accuracy of 97.8%, a recall of 97.3% and an A.P of 97.8%, allowing optimum results in the classification of weeds automatically. |
|---|